Prediction of Microbial Inactivation in Biosolids as aFunction of Moisture Content and Temperature usingArtificial Neural Networks

Agricultural land application of biosolids is an environmentally sound method for recycling stabilized sludge generated during wastewater treatment processes. Meeting regulatory microbial criteria remains an expensive and time consuming process, especially when testing is performed prematurely and requires re-testing. The objective of this study was to determine how well neural network monitoring predicted microbial survival in response to changes in temperature and moisture content of biosolids under laboratory conditions. The backpropagation feed-forward neural network method showed superior performance in predicting the microbial inactivation of fecal coliforms (R2 = 0.95) as a function of temperature and moisture relative to linear-regression models. Although mathematical modeling may not be a sufficient substitute for microbial testing, it may be a useful support tool to optimize the timing of testing for microbial quality designation of biosolids. Further studies involving prediction of specific pathogen levels (Salmonella sp., enteric virus, and viable helminth ova), while incorporating a variety of sludge matrices and environmental parameters, are needed to help determine the full advantages of ANN modeling.